pith. sign in

arxiv: 2604.27849 · v1 · submitted 2026-04-30 · 💻 cs.AI

A Grid-Aware Agent-Based Model for Analyzing Electric Vehicle Charging Systems

Khalil Al-Rahman Youssefi , Marija Gojkovic , Walter Stefanutti , Mika Auer , Melanie Schranz This is my paper

Pith reviewed 2026-05-07 07:34 UTC · model grok-4.3

classification 💻 cs.AI
keywords agent-based modelelectric vehiclescharging systemsgrid integrationsimulationpower managementEV chargingdiscrete-event simulation
0
0 comments X

The pith

An agent-based model simulates electric vehicle charging systems while accounting for grid power limits and user behaviors.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper develops a configurable agent-based model to study electric vehicle charging under different infrastructure setups and operational rules. The model treats each vehicle as an agent with its own charging needs and decisions, while a central Energy Sandbox manages the total power available to prevent overloads. By running simulations of a workplace charging scenario, it reveals how choices like the number of fast versus slow chargers and scheduling methods influence how many vehicles get charged and how the power grid is affected. The work positions this simulation framework as a tool for testing new coordination strategies without requiring immediate real data collection.

Core claim

The paper presents a grid-aware agent-based model implemented in Python with SimPy that integrates heterogeneous EV agent behaviors, charging column constraints, and a shared Energy Sandbox for regulating aggregate power allocation. This enables the analysis of both user-centric charging dynamics and facility-level power behavior in scalable, event-driven simulations. In a representative workplace scenario, the model demonstrates the context-dependent nature of infrastructure suitability, showing that charging strategies and charger types significantly reshape service-level outcomes such as energy delivery and utilization as well as grid-facing characteristics like aggregate load.

What carries the argument

The core mechanism is the Agent-Based Model (ABM) using discrete-event simulation, where individual EV agents interact with constrained chargers under a central Energy Sandbox that enforces power limits on the aggregate load.

If this is right

  • Configuring different numbers and types of chargers affects the overall energy delivered to vehicles and the utilization of infrastructure.
  • Coordination strategies and scheduling rules can improve service outcomes while managing peak power demands on the grid.
  • The model scales to different system sizes, supporting analysis across varying numbers of vehicles and chargers.
  • Results depend on the specific operational context, meaning optimal setups vary by scenario.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Future work could calibrate the agent behaviors against actual charging station data to increase predictive accuracy.
  • This framework could be extended to incorporate renewable energy sources or vehicle-to-grid capabilities.
  • Policy makers might use such models to evaluate the impact of incentives for off-peak charging on grid stability.

Load-bearing premise

The assumption that the simulated heterogeneous EV behaviors and Energy Sandbox power regulations sufficiently capture real-world user decisions and grid dynamics without validation against observed data.

What would settle it

Comparing the model's predicted charging success rates, utilization levels, and aggregate power profiles against measurements from an actual workplace EV charging facility would test if the simulation matches reality.

Figures

Figures reproduced from arXiv: 2604.27849 by Khalil Al-Rahman Youssefi, Marija Gojkovic, Melanie Schranz, Mika Auer, Walter Stefanutti.

Figure 1
Figure 1. Figure 1: ABM of the proposed EV charging framework, view at source ↗
Figure 2
Figure 2. Figure 2: Sequence diagram illustrating the interaction flow view at source ↗
Figure 3
Figure 3. Figure 3: Probability density of TTR9.36 for FCC configurations (Exp. IDs 1, 2, 5, 6, 9, and 10), comparing different EV counts and charging strategies. of charging performance across the EV population and charging scheduling strategy. As observed in the figure, the distribution for 30 EVs is relatively concentrated, while it broadens for 60 and 120 EVs, indicating increased variability in energy delivery times unde… view at source ↗
Figure 4
Figure 4. Figure 4: Cumulative distribution of TTR9.36 for FCC configurations (Exp. IDs 1, 2, 5, 6, 9, and 10), comparing different EV counts and charging strategies view at source ↗
Figure 6
Figure 6. Figure 6: Charging Column (CC) utilization profiles aggregated per configuration. vehicle i; this relation holds even for slow charging columns. Consequently, charging columns remain idle for a substantial portion of the simulation horizon. This observation is structurally tied to the scenario assumptions and does not indicate under-dimensioning of the infrastructure. A second notable effect appears for the configur… view at source ↗
Figure 7
Figure 7. Figure 7: Energy Sandbox power output distribution (15-min bins) for 30 EVs. view at source ↗
Figure 8
Figure 8. Figure 8: Energy Sandbox power output distribution (15-min bins) for 60 EVs. view at source ↗
Figure 9
Figure 9. Figure 9: Energy Sandbox power output distribution (15-min bins) for 120 EVs. view at source ↗
Figure 10
Figure 10. Figure 10: Grid usage for the FCFS strategy, shown together with PV availability and energy price over time in the average employee scenario. price-favorable periods. In the considered workplace scenario, this leads to a temporal mismatch between charging demand and external energy conditions, since charging demand is concentrated early in the day whereas PV availability emerges later. This example therefore highlig… view at source ↗
read the original abstract

This paper presents a configurable, grid-aware Agent-Based Model (ABM) for the systematic analysis of electric vehicle (EV) charging systems under configurable infrastructure and operational conditions. The model integrates heterogeneous EV behavior, charging column constraints, and a shared Energy Sandbox that regulates aggregate power allocation, enabling the joint study of user-centric charging dynamics and facility-level power behavior. Implemented in Python using the SimPy discrete-event framework, the approach supports scalable, event-driven simulations across varying system sizes, charger compositions, and scheduling strategies. A representative workplace charging scenario is investigated to illustrate how infrastructure configuration and coordination mechanisms influence energy delivery performance, infrastructure utilization, and aggregate load characteristics. The results highlight the context-dependence of infrastructure suitability and demonstrate how charging strategies and charger types reshape both service-level outcomes and grid-facing behavior. The proposed ABM provides a flexible and extensible simulation environment for exploring technical, operational, and grid-aware aspects of EV charging ecosystems, and for serving as a methodological basis for subsequent studies on advanced coordination strategies beyond the specific scenario analyzed in this study.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper presents a configurable agent-based model (ABM) implemented in Python with the SimPy discrete-event framework for analyzing EV charging systems. It integrates heterogeneous EV arrival and charging behaviors, charging column constraints, and a shared Energy Sandbox that enforces aggregate power limits. A workplace charging scenario is used to illustrate how infrastructure configurations and scheduling strategies affect energy delivery, utilization, and aggregate load profiles. The authors claim the ABM supplies a flexible, extensible, grid-aware simulation environment that can serve as a methodological basis for studying coordination strategies.

Significance. If the central claims hold, the work supplies a reusable discrete-event simulation platform for exploring interactions between user behavior, charger mixes, and facility-level power constraints in EV systems. The SimPy implementation supports scalability across system sizes, and the scenario results demonstrate context-dependent outcomes that could inform infrastructure design. The approach is standard for ABMs and avoids circularity by using scenario-specific parameters rather than fitted self-referential equations.

major comments (2)
  1. [§3 (Model Architecture, Energy Sandbox subsection)] §3 (Model Architecture, Energy Sandbox subsection): The grid-aware claim rests on the Energy Sandbox enforcing only an aggregate power cap at the facility level. No radial feeder topology, bus voltages, line impedances, unbalanced three-phase power-flow equations, or transformer constraints are modeled. An aggregate limit alone cannot reproduce location-specific effects (voltage drops, phase imbalances, hotspots) that determine real-world charger feasibility, undermining the interpretation of 'grid-facing behavior' and 'infrastructure suitability' in the workplace scenario results.
  2. [§5 (Workplace Scenario Results)] §5 (Workplace Scenario Results): The reported context-dependent outcomes on service-level metrics and load characteristics are presented without sensitivity analysis on free parameters (EV arrival patterns, charger compositions, scheduling thresholds), without error bars or confidence intervals, and without calibration or validation against observed charging data. This leaves the robustness of the performance claims and the positioning of the ABM as a 'methodological basis' unsupported.
minor comments (2)
  1. [Abstract and §1] Abstract and §1: The title and abstract repeatedly use 'grid-aware' and 'grid-facing,' yet the model description clarifies only aggregate power regulation and facility-level behavior. A brief qualification of the term's scope would prevent misinterpretation.
  2. [Implementation] Implementation details: The SimPy event scheduling for heterogeneous EV agents and the exact rule set for the Energy Sandbox power allocation are described at a high level; pseudocode or a small table of decision thresholds would improve reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which help sharpen the scope and presentation of our work. We respond point by point to the major comments, indicating where revisions will be made.

read point-by-point responses

  1. Referee: [§3 (Model Architecture, Energy Sandbox subsection)] The grid-aware claim rests on the Energy Sandbox enforcing only an aggregate power cap at the facility level. No radial feeder topology, bus voltages, line impedances, unbalanced three-phase power-flow equations, or transformer constraints are modeled. An aggregate limit alone cannot reproduce location-specific effects (voltage drops, phase imbalances, hotspots) that determine real-world charger feasibility, undermining the interpretation of 'grid-facing behavior' and 'infrastructure suitability' in the workplace scenario results.

    Authors: We agree that the Energy Sandbox enforces only a facility-level aggregate power cap and does not incorporate detailed distribution-grid elements such as radial topology, voltage calculations, line impedances, or unbalanced power-flow equations. Our use of 'grid-aware' is limited to the incorporation of aggregate power constraints that directly affect charger availability and load profiles within the simulation; this is a standard abstraction for studies focused on operational coordination and utilization at the building or campus scale. We do not claim to reproduce location-specific grid effects. In the revised manuscript we will (i) qualify the term 'grid-aware' explicitly, (ii) restrict references to 'grid-facing behavior' to aggregate load characteristics, and (iii) add a dedicated limitations paragraph that identifies the absence of detailed power-flow modeling as a boundary of the current framework and a natural direction for future extensions that couple the ABM with external power-flow solvers. revision: yes

  2. Referee: [§5 (Workplace Scenario Results)] The reported context-dependent outcomes on service-level metrics and load characteristics are presented without sensitivity analysis on free parameters (EV arrival patterns, charger compositions, scheduling thresholds), without error bars or confidence intervals, and without calibration or validation against observed charging data. This leaves the robustness of the performance claims and the positioning of the ABM as a 'methodological basis' unsupported.

    Authors: The workplace scenario is presented as an illustrative demonstration of the model's configurability rather than a statistically validated prediction exercise. Consequently, the original manuscript did not include systematic sensitivity sweeps or empirical calibration. We accept that this weakens the robustness of the specific numerical outcomes and the strength of the 'methodological basis' claim. In the revision we will add (i) a sensitivity analysis varying key free parameters (arrival-rate distributions, charger-type mixes, and scheduling thresholds), (ii) results aggregated over multiple independent runs with standard-deviation error bars, and (iii) an explicit discussion that positions the current results as exploratory and notes the absence of real-world calibration data as a limitation to be addressed in subsequent studies that apply the model to empirical datasets. revision: yes

Circularity Check

0 steps flagged

No circularity: model is explicitly constructed from standard discrete-event and ABM primitives with scenario parameters

full rationale

The paper defines an agent-based model in SimPy that incorporates heterogeneous EV agents, charger constraints, and an aggregate Energy Sandbox for power regulation. It then executes a configurable workplace scenario to generate illustrative performance metrics. No equations, fitted parameters, or predictions are shown that reduce by construction to the model's own inputs or to self-citations. The central claim is that the implemented simulation environment is extensible; this is a statement about the artifact's design, not a derived result that loops back to hidden assumptions. The absence of empirical calibration is a limitation on external validity, not a circularity in the derivation chain itself.

Axiom & Free-Parameter Ledger

3 free parameters · 2 axioms · 1 invented entities

The central claim rests on several scenario-specific parameters for agent behaviors and infrastructure that are chosen rather than derived, plus domain assumptions about simulation fidelity and the novel Energy Sandbox mechanism introduced without external empirical grounding.

free parameters (3)
  • EV arrival patterns and charging demands
    Heterogeneous user behaviors are parameterized for the workplace scenario without specified fitting to real data.
  • Charger type compositions and power capacities
    Infrastructure parameters are configurable but set specifically for the investigated scenario.
  • Scheduling strategy rules and thresholds
    Coordination mechanisms are defined by the authors for testing different strategies.
axioms (2)
  • domain assumption Discrete-event simulation via SimPy accurately captures the timing and interactions of charging events.
    The framework choice assumes event-driven modeling suffices for the dynamics without continuous-time effects.
  • ad hoc to paper The Energy Sandbox abstraction correctly enforces aggregate power limits without needing real-time external grid data.
    This is a model-specific construct introduced to handle grid constraints.
invented entities (1)
  • Energy Sandbox no independent evidence
    purpose: Regulates aggregate power allocation across chargers to prevent overloads while enabling study of grid-facing behavior.
    New component created for the model; no independent evidence provided beyond the simulation itself.

pith-pipeline@v0.9.0 · 5494 in / 1577 out tokens · 49468 ms · 2026-05-07T07:34:18.459738+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

24 extracted references · 24 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION fin.entry add.period write newline FUNCTION new.block output.state before.all = 'skip after.block 'output.state := if FUNCTION new.sentence output.state after.block = 'skip output.state before.all = 'skip after.sentence 'output.state := if if FUNCTION not #0 #1 if FUNCTION and 'skip pop #0 if FUNCTIO...

    work page

  2. [2]

    Amara-Ouali, Y., Hamrouche, B., Principato, G., and Goude, Y. (2025). Quantifying the uncertainty of electric vehicle charging with probabilistic load forecasting. World Electric Vehicle Journal , 16(2)

    work page 2025

  3. [3]

    Balogun, E., Buechler, E., Bhela, S., Onori, S., and Rajagopal, R. (2024). Ev-ecosim: A grid-aware co-simulation platform for the design and optimization of electric vehicle charging infrastructure. IEEE Transactions on Smart Grid , 15(3):3114–3125

    work page 2024

  4. [4]

    Buss, A. H. and Rowaei, A. A. (2010). A comparison of the accuracy of discrete event and discrete time. In Johansson, B., Jain, S., Montoya-Torres, J., Hugan, J., and Y \"u cesan, E., editors, Proceedings of the 2010 Winter Simulation Conference , Baltimore, MD, USA

    work page 2010

  5. [5]

    N., and Ma, Z

    Christensen, K., J rgensen, B. N., and Ma, Z. G. (2024). Multi-agent based simulation for decentralized electric vehicle charging strategies and their impacts. In Progress in Artificial Intelligence -- EPIA 2024 , volume 14968 of Lecture Notes in Computer Science , pages 220--232. Springer

    work page 2024

  6. [6]

    N., and Ma, Z

    Cong, L., Christensen, K., V rbak, M., J rgensen, B. N., and Ma, Z. G. (2025). Empirically informed multi-agent simulation of distributed energy resource adoption and grid overload dynamics in energy communities. Electronics , 14(20):4001

    work page 2025

  7. [7]

    and Schranz, M

    Gojković, M. and Schranz, M. (2024). Preserving privacy in logistics by using swarm intelligence from the bottom-up. In 2024 IEEE 12th International Conference on Intelligent Systems (IS) , pages 1--7

    work page 2024

  8. [8]

    Holmes, C., Alexander, M., Dunckley, J., Canseco, J., Goldberg, M., Puckett, C., and Williamson, C. (2020). Electric vehicle load shape development for electric utility planning. Technical Update 3002016175, Electric Power Research Institute (EPRI). Prepared in collaboration with DNV GL USA, 122 West Washington Avenue, Suite 1000, Madison, WI 53703-2715; ...

    work page 2020

  9. [9]

    Huaman-Rivera, A., Calloquispe-Huallpa, R., Hernandez, A. C. L., and Irizarry-Rivera, A. (2024). An overview of electric vehicle load modeling strategies for grid integration studies. Electronics , 13(12):2259

    work page 2024

  10. [10]

    R., and Jia, X

    Iranpour, M., Narimani, M. R., and Jia, X. (2025). Assessing ev charging impacts on power distribution systems: A unified co-simulation framework. arXiv preprint

    work page 2025

  11. [11]

    Li, H., Han, B., li, G., Wang, K., Xu, J., and Khan, M. W. (2023). Decentralized collaborative optimal scheduling for ev charging stations based on multi‐agent reinforcement learning. IET Generation, Transmission & Distribution , 18:n/a--n/a

    work page 2023

  12. [12]

    and Obusevs, A

    Luca, G. and Obusevs, A. (2019). Evlpg: Electric vehicle load profiles generator for lv grid studies (software). GitHub repository

    work page 2019

  13. [13]

    R., Thang, K

    Maghami, M. R., Thang, K. F., Mutambara, A. G. O., Firoozi, A. A., Yaghoubi, E., Jahromi, M. Z., and Yaghoubi, E. (2025). Optimized planning of electric vehicle charging infrastructure for grid performance improvement. Discover Sustainability , 6:706

    work page 2025

  14. [14]

    P., Paprocki, M., C ert\' i k, O., Kirpichev, S

    Meurer, A., Smith, C. P., Paprocki, M., C ert\' i k, O., Kirpichev, S. B., Rocklin, M., Kumar, A., Ivanov, S., Moore, J. K., Singh, S., Rathnayake, T., Vig, S., Granger, B. E., Muller, R. P., Bonazzi, F., Gupta, H., Vats, S., Johansson, F., Pedregosa, F., Curry, M. J., Terrel, A. R., Rou c ka, v., Saboo, A., Fernando, I., Kulal, S., Cimrman, R., and Scopa...

    work page 2017

  15. [15]

    and Yip, A

    Muratori, M. and Yip, A. (2024). Projecting electric vehicle electricity demands and charging loads. Technical Report 89775, National Renewable Energy Laboratory (NREL), Golden, CO (United States). OSTI ID: 2367306

    work page 2024

  16. [16]

    Obinata, M., Uchida, H., Yoshizawa, S., Shigematsu, T., Yamaguchi, Y., Shimoda, Y., Ikemoto, Y., Minami, M., and Takeda, K. (2024). Ev charging control using coupled simulation of urban traffic and power distribution system in a real-world area. In uSIM 2024 -- Shaping Net Zero Policies with Building Simulation: The 4th IBPSA-Scotland Conference , Edinburgh, UK

    work page 2024

  17. [17]

    OpenAI (2026). ChatGPT . https://chatgpt.com/. Large language model used for language refinement and editing

    work page 2026

  18. [18]

    Pedrielli, G., Tolio, T., Terkaj, W., and Sacco, M. (2012). Distributed modeling of discrete event systems. In Distributed Modeling of Discrete Event Systems , chapter 1. InTechOpen

    work page 2012

  19. [19]

    Plagowski, P., Saprykin, A., Chokani, N., and Shokrollah-Abhari, R. (2021). Impact of electric vehicle charging -- an agent-based approach. IET Generation, Transmission & Distribution , 15

    work page 2021

  20. [20]

    Schranz, M., Harshina, K., Forg \'a cs, P., and Buining, F. (2024). Agent-based modeling in the edge continuum using swarm intelligence. In Proceedings of the International Conference on Agents and Artificial Intelligence (ICAART 2024) , Rome, Italy

    work page 2024

  21. [21]

    Introduction to discrete-time simulation

    Software Solutions Studio (2022). Introduction to discrete-time simulation. https://softwaresim.com/blog/introduction-to-discrete-time-simulation/. Accessed: 2026-03-02

    work page 2022

  22. [22]

    Xu, Q., Wu, Z., Xin, S., Niu, H., Zhang, P., and Wang, X. (2025). Grid capacity planning model for electric vehicle high charging penetration in power distribution networks. International Journal of Low-Carbon Technologies

    work page 2025

  23. [23]

    A., Gojkovi \' c , M., and Schranz, M

    Youssefi, K. A., Gojkovi \' c , M., and Schranz, M. (2025). Enhancing job-shop scheduling performance with a refined bottom-up variant of the artificial bee colony algorithm. In Wagner, G., Werner, F., and De Rango, F., editors, Simulation and Modeling Methodologies, Technologies and Applications , pages 94--113, Cham. Springer Nature Switzerland

    work page 2025

  24. [24]

    A.-R., Gojković, M., Stefanutti, W., Auer, M., and Schranz, M

    Youssefi, K. A.-R., Gojković, M., Stefanutti, W., Auer, M., and Schranz, M. (2026). Agent-based model for the sharedcharging project. DOI: doi:10.5281/zenodo.18841747

    work page doi:10.5281/zenodo.18841747 2026